【摘要】：梯度向量流(GVF)有效解決了主動輪廓(snakes)模型初始化和凹陷區(qū)域收斂的問題,但由于其各向同性的擴散特性,使得對弱邊緣和角點的捕獲能力不足。因此,致力于尋求一種GVF各向異性擴散機制。通過將拉普拉斯算子進行正交分解,分析了GVF模型的法向和切向擴散作用,發(fā)現(xiàn)(類似)角點處的GVF場存在明顯的曲率收縮和切向退化,進一步揭示了角點和弱邊緣丟失的原因。在此基礎(chǔ)上,通過對法向GVF(NGVF)模型引入新的邊緣保真項和有偏的權(quán)重系數(shù),提出一種新的外力模型。最后,通過實驗對該方法的分割準(zhǔn)確性和計算效率進行了比較分析。實驗結(jié)果表明,該方法在保持一定計算優(yōu)勢同時,能準(zhǔn)確地捕獲弱邊緣和角點。
[Abstract]:Gradient vector flow (GVF) effectively solves the problem of initializing the active contour (snakes) model and converging the concave region, but because of its isotropic diffusion property, the capture ability of weak edges and corners is insufficient. Therefore, we are looking for a GVF anisotropic diffusion mechanism. By orthogonal decomposition of Laplace operator, the normal and tangential diffusion of the GVF model is analyzed. It is found that the GVF field at the corner point has obvious curvature contraction and tangential degeneration. The causes of the loss of corners and weak edges are further revealed. On this basis, a new external force model is proposed by introducing new marginal fidelity terms and biased weight coefficients to the normal GVF (NGVF) model. Finally, the segmentation accuracy and computational efficiency of the method are compared and analyzed by experiments. The experimental results show that the method can accurately capture the weak edges and corners at the same time.
【作者單位】： 火箭軍工程大學(xué)一系;
【分類號】：TP391.41

【相似文獻】

相關(guān)期刊論文 前10條

1 何源;羅予頻;胡東成;;基于梯度向量流的醫(yī)學(xué)圖像自動分割[J];計算機應(yīng)用;2007年01期

2 王志宏;袁Y，

本文編號：2199758